what kind of teachers are schools looking for? evidence ... · rothstein, john yinger; seminar...

1

What Kind of Teachers Are Schools Looking For? Evidence from a Randomized

Field Experiment*

Peter Hinrichs∗∗

First Draft: March 1, 2013 This Draft: November 20, 2013

Abstract: Teacher quality is a pressing public policy concern, yet there is little evidence

on what types of teachers schools actually prefer to hire. This paper reports the results of

an experiment that involved sending schools fictitious resumes with randomly-chosen

characteristics in an attempt to determine what characteristics schools value when hiring

new teachers. The results of the study suggest that an applicant’s academic background

has little impact on the likelihood of success at private and charter schools, although

public schools respond more favorably to candidates from more selective colleges.

Additionally, private schools demonstrate a slight preference for female candidates, and

all three sectors demonstrate a preference for in-state candidates.

Keywords: resume audit studies, teacher labor markets JEL Classification: I21, J45

* I thank Duncan Chaplin, Thomas DeLeire, John Friedman, Nora Gordon, Fabian Lange, Jonah Rockoff, Jesse Rothstein, John Yinger; seminar participants at Georgetown University, Georgia State University, and Texas A&M University; lunch seminar participants at the Federal Reserve Board; and participants at the 2013 Association for Education Finance and Policy conference, Association for Public Policy Analysis and Management conference, National Bureau of Economic Research spring economics of education program meeting, and Society of Government Economists conference for helpful discussions and comments. I thank Malini Silva and Robert Wood for research assistance. ∗∗ McCourt School of Public Policy, Georgetown University, 100 Old North Building, 37th & O Streets NW, Washington DC 20057. E-mail: [email protected].

2

I. Introduction

A growing body of empirical work documents the enormous heterogeneity in quality

amongst classroom teachers.1 This quality matters. Recent work by Chetty, Friedman, and

Rockoff (2011) finds that teacher quality in the early grades affects students’ earnings in

adulthood. And Hanushek (2011) writes, “Replacing the bottom 5-8 percent of teachers with

average teachers could move the U.S. near the top of international math and science rankings

with a present value of $100 trillion.” It follows from this line of work that the distribution of

teacher quality has the potential to dramatically affect the level and distribution of national

income.

Unsurprisingly, there is a concerted effort to raise teacher quality. Programs like Teach

for America have the goal of encouraging academically-talented recent college graduates to

become teachers, and merit pay policies that are in place in some districts aim to have

individuals who would be effective teachers select into the teaching profession. However, these

policies and programs generally focus on the supply side of teacher labor markets. Less attention

has been paid to the demand side. But clearly, the policies and actions of the districts, schools,

and administrators on the hiring side of teacher labor markets affect who becomes a teacher as

well. If these actors are hiring teachers suboptimally, there may be a potential to raise teacher

quality by simply making changes to the hiring process. But despite this potential, little is

known about how effective schools are in screening applicants or about what characteristics they

seek in potential teachers.2

1 Seminal work includes Aaronson, Barrow, and Sander (2007); Hanushek (1971); Murnane (1975); Rivkin, Hanushek, and Kain (2005); and Rockoff (2004). But also see Rothstein (2009, 2010) for an influential critique of conventional estimators of teacher value-added. 2 There are some previous studies, such as Harris et al. (2010), which survey principals about what kinds of teachers they are looking for, although the sample sizes are typically small and it is not clear that actual hiring behavior is consistent with responses to surveys.

3

This paper sheds some light on the demand for teacher characteristics through a

randomized controlled experiment in the labor market for new teachers. The experiment

randomly manipulates characteristics on resumes submitted by fictitious candidates for teaching

positions and then studies how the responses to different characteristics vary. In particular, I sent

6,000 fictitious resumes to randomly-selected schools across the United States, along with cover

letters expressing an interest in being hired for a teaching position. The resumes attempt to

experimentally induce the demand side’s perceptions of a candidate’s academic credentials, sex,

geographic location, and other characteristics. Due to the random assignment of these resume

characteristics, comparing responses to the various resumes should provide a credible estimate of

what characteristics schools value in the initial screening stage when hiring new teachers. The

results of the study suggest that an applicant’s academic background has little impact on the

likelihood of success at private and charter schools, although public schools respond more

favorably to candidates from more selective colleges. Additionally, private schools demonstrate

a slight preference for female candidates, and all three sectors demonstrate a preference for in-

state candidates.

The rest of this paper is organized as follows: Section II provides relevant background

information, Section III discusses the methodology used in this experiment on teacher labor

markets, Section IV gives the results, and Section V concludes.

II. Background Information

A. Previous Research on Teacher Hiring

Earlier research on teacher labor markets includes work on teacher labor supply (Bacolod

2007; Corcoran, Evans, and Schwab 2004a, 2004b; Engel and Jacob 2011; Ransom and Sims

4

2008), the sorting of teachers across schools (Boyd et al. 2005a; Clotfelter, Ladd, and Vigdor

2005, 2006; Goldhaber, Gross, and Player 2011; Jackson 2009; Lankford, Loeb, and Wyckoff

2002), and the impact of counterfactual personnel policies on teacher quality (Rothstein 2012;

Staiger and Rockoff 2010). There is less work on how effective schools are in screening

teachers. However, Kane and Staiger (2005) find that teachers hired as part of a hiring surge by

Los Angeles Unified School District to comply with a California class size reduction policy did

not perform significantly differently from teachers hired the previous year as part of a much

smaller cohort. This provides indirect evidence that the district was not effective in screening

teachers because, if it were, one would expect the marginal teacher to be worse than the average

teacher and for average teacher quality to fall as more teachers are hired.3

More directly related to the present study is Ballou (1996), an influential paper that

addresses the question of how interested schools actually are in hiring academically-talented

teachers. Ballou (1996) uses data from several waves of the Survey of Recent College Graduates

and finds that, of those individuals who applied for any teaching position, those who had more

impressive academic qualifications in terms of college selectivity or having majored in math or

science were not more likely to be later found working as a teacher than those with less

impressive academic qualifications. Ballou (1996) interprets this result as showing that the

demand side does not show much interest in hiring teachers who are academically strong. But

although Ballou (1996) provides additional information that supports this interpretation of the

results, it is difficult to completely rule out the possibility that the results are driven by applicants

who are more academically talented having better outside options than those who are less

academically talented. They may thus have lower search intensities or be less likely to accept a

3 Also see Jepsen and Rivkin (2009), a statewide study of California’s class reduction policy that finds “little systematic relationship between cohort size and teacher quality.”

5

position once offered, which could explain why talented applicants who applied for at least one

teaching position do not end up working as teachers without implying anything about the

preferences of schools over applicants for any particular teaching position.

Boyd et al. (2011) attempt to circumvent some of the difficulties of Ballou (1996) by

employing data on teachers’ applications to transfer to specific schools in New York City. These

authors find that, of those teachers who applied for a transfer, those who had higher certification

exam scores, who had a higher value-added, and who attended more selective colleges were

more likely to be working in the new position the following year. Boyd et al. (2013) obtain

similar results when using data on the matching of teachers to jobs in New York State to estimate

a structural, game-theoretic, two-sided matching model of the teacher labor market. The general

results in Boyd et al. (2011) and Boyd et al. (2013) are thus in contrast to Ballou (1996). The

present paper thus seeks to cleanly identify preferences of schools over candidates at the initial

screening stage by randomly assigning academic qualifications to resumes, sending them to

specific schools, and then monitoring the responses received from these particular schools.

There is also a small body of research on the narrow geographic scope of teacher labor

markets. A survey of Pennsylvania school superintendents found that, in the average district,

40% of the teachers had previously attended high school within the district (Strauss et al. 2000).

Boyd et al. (2013) note that, “In New York State, over 60% of teachers first teach within 15

miles of the high school from which they graduated and 85% teach within 40 miles.” And,

Reininger (2012) finds, “Across the country, the median distance moved by teachers [relative to

where they lived in 10th grade], 13 miles, is much less than that of other college graduates, 54

miles, and is more similar to the median distance used by high school graduates, 7 miles.”

6

These statistics on geographic mobility may suggest that schools are casting a narrow net

when searching for teachers. If this is true, it may be problematic because a broader search may

result in better candidates. However, an alternative explanation for these statistics is that the

candidates themselves may be particularly interested in working near where they grew up. The

results of the matching model in Boyd et al. (2013) suggest that teachers do have a preference to

work at nearby schools but also that schools do in fact also have a preference to hire teachers

who live near the school at the time they applied for certification. Killeen, Loeb, and Williams

(2013) obtain similar results when studying job application data for teachers in Vermont.4 The

present study builds on this earlier work by randomly varying the stated geographic location of

job seekers in an attempt to cleanly identify the extent to which the demand side is responsible

for the narrow geographic scope of teacher labor markets.

Although little is known about gender discrimination in teacher hiring, research suggests

that labor market discrimination against women exists in other contexts (Altonji and Blank 1999;

Goldin and Rouse 2000). But one interesting feature of the teaching profession is that it is one in

which women are overrepresented. According to tabulations from the Current Population

Survey, only 18.6% of elementary and middle school teachers are men. Men are even

underrepresented as high school mathematics teachers, as tabulations from the Schools and

Staffing Survey reveal that only 43.2% of mathematics teachers in grades 9-12 at public schools

are men.5 A goal of this study is to determine whether schools themselves have a role in

exacerbating or reversing these disparities.

4 Also see Boyd et al. (2005) on the relationship between geographic proximity and teacher turnover. 5 The elementary and middle school figure is for 2012 and was found at http://www.bls.gov/cps/cpsaat11.pdf (accessed February 24, 2013). The high school mathematics figure is for 2007-08 and can be found in Table 75 of the 2012 Digest of Education Statistics.

7

B. The Relationship between Teacher Characteristics and Teacher Quality

An effective teacher hiring process would distinguish among candidates along

dimensions that are related to productivity but would not discriminate based on irrelevant

characteristics. This raises the question of whether the characteristics I consider in this study

actually are related to teacher quality. For example, one might contend that it is unproblematic

that schools treat applicants of high academic ability and low academic ability equally if

academic ability is unrelated to success as a teacher.

Although there is not a universal consensus on the topic, there is a strong case to be made

that academically-talented teachers are in fact better in the classroom. Hoxby and Leigh (2004)

state that it is a matter of “logic” that “a teacher’s value-added is related to her academic

aptitude.” Ballou and Podgursky (1997) argue that, “The link between teachers’ cognitive

abilities and student learning stands out in a literature that frequently fails to find significant

relationships between other teacher attributes and student achievement.”6 A review article by

Goldhaber (2008) notes that, although teacher quality is not generally associated with easily-

observed characteristics of teachers, “Some readily identifiable characteristics do predict success

in the classroom. In particular, measures of academic proficiency or cognitive ability, such as a

teacher’s performance on standardized tests (e.g., licensure tests or the SAT or the selectivity of

the colleges she graduated from), and subject specific training (e.g., a degree in mathematics) in

a teacher’s specialty area appear to be predictors of teacher quality.” Similar results have been

found by a more recent wave of studies, including Clotfelter, Ladd, and Vigdor (2010);

6 Another exception to the general result that observable teacher characteristics are not associated with achievement is that research generally finds that teachers improve over their first couple years on the job (see, e.g., Goldhaber 2008). However, experience is not a characteristic that I vary in this study because it may be easier for school officials to determine that a resume from an experienced applicant is fictitious. Thus, all the resumes in this study are for new teachers.

8

Goldhaber (2007); and Jackson and Bruegmann (2009).7 This perceived relationship between

between teacher aptitude and student success has also apparently motivated organizations such as

Teach for America, which has the stated mission of “eliminat[ing] educational inequality by

enlisting high-achieving recent college graduates and professionals to teach for two or more

years in high potential communities throughout the United States.”

Additional evidence in favor of the proposition that academically-talented teachers are

better teachers comes from recent studies of Teach for America. One study based on random

assignment of students to teachers finds that Teach for America teachers have a positive effect

on math test scores, albeit not on reading test scores (Glazerman, Mayer, and Decker 2006). An

observational study by Xu, Hannaway, and Taylor (2011) also finds a positive effect of Teach

for America teachers on test scores. Also of note is Dobbie (2011), who finds that, amongst

Teach for America teachers, those who are rated as having high academic achievement have a

positive effect on student math test scores.

Researchers have also studied the effects of teacher gender. Work by Dee (2005, 2007)

suggests that there are academic benefits when students are matched with a teacher of the same

gender as themselves. Dee (2007) finds that, excluding mathematics, the test score gain to girls

from having a female teacher is roughly equal to the test score loss to boys from having a female

teacher. Thus, if the students in a classroom are balanced on gender, we would not expect

teacher gender to have an effect on average achievement in these subjects even though it might

affect the distribution of achievement.8

7 However one recent example to the contrary is Harris and Sass (2011), which finds that teachers’ SAT scores are not associated with teacher value-added. 8 However, Dee (2007) also finds that having a female teacher has a negative relationship with math test scores for both boys and girls, although he suggests that this result may be due to nonrandom sorting of teachers to classrooms. But Antecol, Eren, and Ozbeklik (forthcoming) find, based on random assignment of students to teachers, that having a female teacher actually is associated with lower math test scores for female students. To the extent that this is true, having more male teachers may raise achievement across the board.

9

However, the results about gender match between students and teachers are noteworthy

given that males now lag behind females on a variety of academic outcomes, including college

attendance rates (Jacob 2002; Goldin, Katz, and Kuziemko 2006). Having a higher percentage

of male teachers may reduce these gender gaps. Some commentators, such as Gormley (2012),

have thus called for a larger number of male teachers. Although one may argue that male

underperformance in school may not be a problem because men still outperform women in the

labor market, it is also worth keeping in mind that men also fare worse than women on a number

of “left tail” outcomes and that education may mitigate this problem. For example, Lochner and

Moretti (2004) and Deming (2011) both note that men commit much more crime than women

and also find that schooling has the potential to reduce crime.

Finally, the relationship between teacher geographic proximity and student achievement

is unclear.9 However, insofar as a preference for nearby applicants is indicative of schools

casting a narrow net for potential teachers, this would be expected to result in worse hiring

decisions. Furthermore, as with the other variables considered in this study, it is still of interest

to know which characteristics schools actually do consider in hiring even if there is not universal

consensus on which characteristics they ought to consider. This can further our understanding of

how teacher labor markets operate and add to our knowledge about the composition of the

teaching workforce.

C. Resume Audit Studies

The practice of studying the responses to fictitious job applications in order to measure

employer preferences is known as a “resume audit study” or a “correspondence study.” This

9 Although some may argue that candidates from nearby will better understand the unique local context, it is not clear how important this is. Furthermore, one could just as easily make the case that students will benefit from being exposed to teachers who are different from themselves.

10

methodology was employed as early as 1970 to test for discrimination against immigrants in

England (Jowell and Prescott-Clarke 1970). The methodology has recently enjoyed increased

popularity in economics, owing in large part to Bertrand and Mullainathan’s (2004) study of

whether employers discriminate against job applicants with distinctively black names. Resume

audit studies have also been used to study discrimination based on age (Lahey 2008), gender

(Riach and Rich 2006), sexual orientation (Weichselbaumer 2003), immigrant status

(Oreopoulos 2011), and obesity (Rooth 2009). Recent resume audit studies have gone beyond

studying whether employers discriminate based on demographic and physical characteristics to

study such topics as the extent to which employers value mathematics skills (Koedel and Tyhurst

2012) and how employers weigh unemployment spells of various durations (Kroft, Lange, and

Notowidigo 2013).

A strength of resume audit studies is that they provide the researcher control over all

information employers can observe about a candidate. This allows the researcher to randomly

assign resume characteristics and isolate the effects of these characteristics on employer

responses. This overcomes some of Heckman and Siegelman’s (1993) criticisms of in-person

audit studies, such as the possibility that the testers will differ from one another along important

unobservable dimensions and the possibility that the testers will act in a way that leads to the

results they believe the experimenter wants to find.10

However, a limitation of the typical resume audit study is that the researcher can observe

only whether or not a candidate is called in for an interview, which may not provide a complete

picture of the hiring process. Nonetheless, Riach and Rich (2006) point out, based on studies

that send out fictitious resumes and then follow up with interviews of trained actors posing as job

10 Some prominent examples of in-person audit studies are Neumark, Bank, and Van Nort (1996); Ondrich, Ross, and Yinger (2003); and Yinger (1986).

11

seekers, that most discrimination takes place at the initial resume screening stage of the hiring

process. Thus, studying this initial screening stage seems to provide an effective means of

gauging employer preferences. Intuitively, if hiring personnel have preferences over easily-

observed characteristics such as gender, age, or academic credentials, it would seem that they

would be able to exercise those preferences early on in the hiring process. Furthermore,

although who is hired is likely of more interest than who is interviewed, there is a relationship

between the two in that the pool of interviewees is also presumably the pool of potential hires.

Thus, even if hires are made randomly from the pool of interviewees, factors that affect the

probability of receiving an interview should also affect the unconditional probability of being

hired.

III. Methods

A. Selecting the Sample

The first step of this resume audit study was to select the schools involved. I selected

3,000 schools to receive two resumes each. Thus, the overall sample of 6,000 is similar to that in

earlier resume audit studies, such as Bertrand and Mullainathan (2004) and Lahey (2008).11 The

schools were selected at random from the 2009-10 Common Core of Data, which includes data

on charter schools in addition to traditional public schools, and the 2009-10 Private School

Survey. These data sets are intended to form a complete census of schools in the United States,

and the 2009-10 data were the most recent data available at the time of the study. In order to

explore heterogeneity across school sectors, the sample consists of 1,000 traditional public

11 This sample size is also supported by power calculations and a small pilot study I conducted. Details are available upon request.

12

schools, 1,000 charter schools, and 1,000 private schools. Within each of these sectors, schools

were sampled without replacement with a probability proportional to student enrollment.12

B. Creating the Resumes

The next step of the study was to create the fictitious resumes. The goal was to create

realistic-looking resumes for recent college graduates seeking their first teaching position. To

aid in this process I consulted guidebooks for prospective teachers, as well as some actual

resumes of current and former teachers.13 I then created one-page resume templates that were

similar in style to the actual resumes I consulted. The resume templates contain fields to fill in a

candidate’s name and contact information, information on the candidate’s educational

background and licensure status, a list of personal strengths, and information on student teaching

and other previous work experience. I use a variety of values for each of these variables, which

may help overcome Heckman and Siegelman’s (1993) critique of previous audit studies that

estimate discrimination at only a single value of the background characteristics.

Characteristics of the fictitious job applicants were generally filled in to the resume

templates at random and independently from one another, but an exception is the information on

college major and teacher licensure. Based on conversations with officials at state licensure

agencies in a number of states, all resumes sent to elementary schools listed a major and

certification in elementary education. All resumes sent to secondary schools list a major and

certification in mathematics, and with probability .25 the secondary school resumes list an

12 In a small number of cases in which multiple schools in the sample had the same principal or administrator, one of the schools was selected at random to remain in the study and the rest were replaced by a new school selected at random from the relevant population. 13 These guidebooks include Anthony and Roe (2003); Brause, Donohue, and Ryan (2002); Clement (2007); Enelow and Kursmark (2011); Feirsen and Weitzman (2004); Hougan (2011); McKinney (2000); Pollock (2011); Warner, Bryan, and Warner (2006); and Wei (2010).

13

additional certification in science. These fields were chosen in an attempt to maximize power for

a given sample size, on the belief that job applicants in math and science would be more likely to

receive a positive response relative to those in other disciplines.

The main academic credentials I consider in this study are grade point average (GPA) and

college attended. Resumes were assigned a grade point average of 3.1, 3.5, and 3.9 with

probability 1/3 each. The procedure for choosing the colleges is slightly more complicated. I

began by randomly assigning each resume to list either a college in the same state as the school

the resume was to be sent to (with probability .75) or a college in a different state (with

probability .25).14 I then selected all colleges in the 2011 edition of Barron’s Profiles of

American Colleges that offered majors in both elementary education and mathematics. Barron’s

assigns colleges to nine quality tiers, and not every state has a college in each quality tier. The

in-state resumes were given a college in the highest selectivity tier of colleges in the state, a

college in the lowest selectivity tier of colleges in the state, and a college in one of the middle

selectivity tiers each with probability 1/3.15 The out-of-state resumes were assigned a college in

a similar manner, except that the three selectivity categories were based on colleges nationwide

rather than just those in a particular state.

The names of the fictitious applicants were selected at random from names that were

popular at the time the applicants likely would have been born. I utilized the five most common

14 Importantly, schools receiving out-of-state resumes is not a rarity. For example, Killeen, Loeb, and Williams (2013) find that about 45% of applicants for teaching positions in Vermont are from outside the state. Although this figure is likely higher for Vermont than other states due to its small size, the point is that it is possible for teachers to cross state lines. 15 For example, in Florida there are 23 institutions that meet the requirement of offering both a mathematics major and an elementary education major. The highest rated of these was The University of Miami, which falls in the “most competitive” category. Thus, all the resumes from the Florida in-state sample that were selected to have the highest selectivity level list The University of Miami as the college attended. The Florida in-state resumes that were selected to have the lowest selectivity level list one or another of the three institutions in Florida that offer both a mathematics major and elementary education major and are rated by Barron’s as being “less competitive.” The middle selectivity resumes from Florida list one or another of the 19 remaining universities in Florida that offer both a mathematics major and an elementary education major. Each college that matches the state and selectivity tier the college is to be selected from was equally likely to be chosen.

14

last names in the 1990 census (Brown, Johnson, Jones, Smith, and Williams), the ten most

common first names for girls born in 1990 (Amanda, Ashley, Brittany, Elizabeth, Jennifer,

Jessica, Lauren, Samantha, Sarah, and Stephanie), and the ten most common first names for boys

born in 1990 (Andrew, Christopher, Daniel, David, James, Joseph, Joshua, Justin, Matthew, and

Michael).16 The study uses all 100 combinations of first and last names amongst these popular

names. With the assistance of a direct mail marketing company, the resumes were randomly

assigned actual apartment addresses in or near the city that the college listed on the resume is

located in. Although I was not able to monitor any responses received by U.S. mail, one

previous audit study that was able to do so found that very few employers responded by U.S.

mail; moreover, when they did respond, it was never to request an interview (Lahey 2008). Each

resume also lists student teaching experience at a school selected at random from the Common

Core of Data that is in or near the city in which the applicant’s college is located. The resumes

were also randomly assigned additional previous work experience, as well as a list of personal

strengths. Finally, the resumes were given functioning e-mail addresses and phone numbers in

order to monitor the responses.

C. Sending the Resumes

One way in which this study differs from previous resume audit studies is that this study

sends unsolicited e-mails to school administrators rather than applying to posted job openings.

Although this was done for practical reasons, it is worth noting that Heckman and Siegelman

(1993) criticize the practices of previous resume audit studies on the grounds that many job

openings are not actually posted. Applying for only posted positions may therefore potentially

16 The first names come from http://www.ssa.gov/cgi-bin/popularnames.cgi, and the last names come from http://www.census.gov/genealogy/www/data/1990surnames/names_files.html.

15

result in misleading measures of employers’ preferences over candidates. Additionally,

according to guidebooks for prospective teachers, sending unsolicited resumes is a recommended

method of searching for a teaching position (Brause, Donohue, and Ryan 2002; McKinney 2000;

Wei 2010).17 Finally, the reasonably high response rate the unsolicited resumes received from

schools in this study validates this method of job search.

Each school in this study received two resumes, generally one in June 2012 and one in

August 2012. The purpose of the two-month lag between resumes is to lessen any suspicion of

the two resumes having the same origin. Furthermore, each school received a resume in the

second round that used a different format and style than the one it had received in the first round.

All resumes were accompanied by a brief cover letter expressing an interest in being interviewed

for a teaching position. The resumes were generally sent by e-mail to the principal, headmaster,

or other lead administrator of the school.18 I obtained e-mail addresses of school administrators

by searching through state directories, looking at school websites, and calling schools and

directly asking for the principal’s e-mail address without providing any information about the

purpose of the study. Resumes were sent by US mail to schools for which I was unable to obtain

the head administrator’s e-mail address using one of these three methods.

Finally, a word about the timing is in order. According to the guidebooks for prospective

teachers I consulted, the market for new teachers occurs over an extended period of time but

many hiring decisions are not made until just before the school year begins. One guidebook

states, “May and June are the busiest months for hiring teachers….Hiring activity slows in

17 The formal hiring process and the amount of discretion the principal has vary across school districts, although Rutledge et al. (2008) explain that principals can find ways to circumvent the formal process even in cases in which the rules make it difficult for them to hire their preferred candidates. 18 Due to an apparent glitch with an e-mail add-in, for a small number of e-mails there is no record in the “sent items” folder of the e-mail actually having been sent. In these cases, I resent the e-mail. The main results are robust to alternative treatments of these cases, including controlling for these cases with a dummy variable or dropping them from the sample. Additionally, due to human error, a small number of e-mails were sent from a different e-mail address than originally intended. The results are also robust to alternative treatments of these cases.

16

July….Hiring picks back up in August and September as principals try to fill remaining

vacancies, as well as last minute teacher transfers and retirements” (Hougan 2011, p. 140).19

The four large urban districts studied in a recent report by the New Teacher Project all still had

vacancies after the school year has begun (Levin and Quinn 2003). Moreover, Engel’s (2012)

tabulations of data from the Schools and Staffing Survey suggest that 25% of new teachers are

hired before the previous school year ends, 30% are hired during the first half of the summer,

34% are hired during the second half of the summer, and 11% are hired after the school year has

already begun. In an attempt to send resumes at around the time schools would be hiring, I opted

to send the resumes in June and August. The resumes sent in August had roughly the same

response rate as those sent in June.

D. Coding the Responses

I monitored the e-mail addresses and voicemails for responses until September 30, 2012.

I then coded variables based on the type of response received. The main outcome variables

employed in this study are a dummy for whether a resume received an interview request and a

dummy for whether a resume received either an interview request, a request for more

information, or a request to apply for or interview for a different position (e.g., a substitute

teaching position.) I thus follow earlier authors, such as Lahey (2008), by considering both a

broader and a narrower measure of success.

E. Models and Estimators

19 Also see Feirsen and Weitzman (2004). Moreover, a survey of New York State school superintendents conducted by Balter and Duncombe (2008) finds that the average school district typically makes job offers in June. See Papay et al. (2013) on the consequences of late hiring.

17

Due to the random assignment of the resume characteristics, the analysis of the data is

relatively simple and straightforward. Simple differences of means should provide unbiased

estimates. I also show results of regression models estimated by ordinary least squares. 20 The

full specification is

+⋅+⋅+⋅+= ijsijsijsijs ivityhighselectmediumGPAhighGPApositive 3210 ββββ

ijssijsijsijsijs xoutofstatefemalectivitymediumsele εδββββ +++⋅+⋅+⋅ 7

'

654 .

The unit of observation is a resume, with resume i being sent to school j in state s . Here

ijspositive is an indicator for receiving a positive response to a resume, ijshighGPA is an

indicator for the resume listing a GPA of 3.9, ijsmediumGPA is an indicator for the resume listing

a GPA of 3.5, ijsivityhighselect is an indicator for the resume belonging to the high selectivity

tier, ijsctivitymediumsele is an indicator for the resume belonging to the medium selectivity tier,

ijsfemale is an indicator for the resume listing a female name, and ijsoutofstate is a dummy for

the resume listing an address and college from outside the state the receiving school is in.

Control variables for whether the resume lists an additional certification in science, the level of

the school, the sector of the school, the racial composition of the school, and the urbanicity of the

school are included in the vector ijsx . The term sδ denotes a full set of state indicators for the

receiving school. The error term is ijsε , and the β ’s are parameters to be estimated. I report

standard errors that are clustered at the school level. I also estimate some models that do not

control for covariates, including some that enter the various treatments in isolation. However,

this does not have much impact on the point estimates, which is unsurprising given that the

resume characteristics are assigned randomly. The effects on the standard errors are minimal as

20 Although the results shown in this paper are from linear probability models, the results are very similar when estimating probits and logits.

18

well. On the other hand, there are substantive changes in the results when I reweight the sample

(which consists of an equal number of public schools, charter schools, and private schools) to be

representative of the school the average student is attending. Finally, I also explore

heterogeneity by school sector and between elementary and secondary schools.

IV. Results

Table 1 shows summary statistics. All variables except for the “Fraction

Underrepresented Minority” variable are binary, so only means are shown in the main body of

the table. The table indicates that roughly 4.3% of resumes sent received an interview request.

When I define “positive outcome” more broadly to include cases in which the school asked for

additional information about the candidate or asked the candidate to apply or interview for a

different position the rate of positive response is roughly 7.9%. These figures are lower than the

corresponding figures in previous resume audit studies, which is not altogether surprising given

that the resumes in this experiment were not sent in response to posted positions. What is

perhaps more surprising is that these figures are not too much lower than the corresponding

figures in other studies. For example, Lahey (2008) and Kroft, Lange, and Notowidigdo (2013)

both obtain interview request rates of about 4.7%, and under Lahey’s broader definition of

“positive response” the success rate is about 9.0%. The remaining rows of Table 1 show

characteristics of the schools that received the resumes and also demonstrate that the actual

assignments of the treatment variables are similar to the intended probabilities. The right-hand

column of Table 1 shows means after reweighting the three sectors by relative enrollment. It is

clear from examining this column that enrollment in traditional public schools (88.9%) is much

higher than enrollment in either private schools (8.1%) or charter schools (3.0%). Furthermore,

19

public schools are disproportionately located in rural areas and are less likely to give a positive

response or an interview request in response to receiving a resume.

The main results are first presented in Table 2. This table shows the raw positive

response rate and the raw interview request rate for the various treatment conditions. The table

shows that only 7.4% of candidates listing a high GPA of 3.9 received a positive response,

compared to 7.9% of those listing the medium GPA of 3.5 and 8.3% of those listing the low

GPA of 3.1. This pattern remains the same, although the magnitudes diminish, after reweighting

by enrollment in the three sectors. Additionally, the resumes in the study listing a more selective

college were slightly more likely to receive a positive response than those listing a less selective

college. This gap widens when reweighting by enrollment, which suggests that there are

differential effects by sector. Those listing a female name were slightly more likely to receive a

positive response than those listing a male name, although this pattern is reversed when

reweighting. Those that listed an in-state college and address were much more likely to receive a

positive response than those that listed an out-of-state college and address, and this pattern

remains after reweighting the sample. Lastly, although the magnitudes are naturally lower for

interview requests, the table shows similar qualitative results for interview requests as for

positive responses. This suggests that the “positive response” variable and the “interview

request” variable may be measuring a similar underlying construct.

Table 3 shows the results for the “positive response” variable in a regression context.

The table shows estimates for both the unweighted and the reweighted sample. The first four

columns of each side of this table consider each of the four main sets of treatment variables in

isolation, the fifth column combines all of the treatments into one regression, the sixth column

adds additional covariates, and the seventh column includes a full set of dummy variables

20

identifying the state the receiving school is located in. Consistent with the random assignment of

resume characteristics, the coefficients on the treatment variables do not change much across

columns within these seven columns. However, some of the results do change between the

unweighted sample and the weighted sample, which points to heterogeneous effects by sector.

The results of Table 3 suggest that having a high GPA does not help candidates. In fact,

all of the point estimates are negative, albeit not statistically different from 0. The fact that

having a low GPA is not an impediment to receiving a positive response is particularly

noteworthy given the level of grade inflation in education schools documented by Koedel (2011).

A GPA of 3.1 is potentially very low in the distribution, so it is interesting that these resumes are

treated similarly to those listing a GPA of 3.5 or 3.9. Moreover, although the general results of

Ballou (1996) suggest that strong academic qualifications do not help in obtaining a teaching

position, Ballou (1996) actually finds a positive effect of undergraduate grade point average.

However, although the effects of college selectivity are not statistically significant in the

unweighted sample, they are large in magnitude and highly significant in the weighted sample.

The results of column 14 suggest that attending a moderately selective college rather than a

college in the lowest selectivity tier is associated with a 1.84 percentage point increase in

positive responses, and attending a college in the highest selectivity tier is associated with a 3.61

percentage point increase in positive responses. By way of comparison, column 7 suggests

statistically insignificant 0.82 percentage point and 1.29 percentage point increases. The

difference across samples shows that there are differential effects across sectors. In particular,

there is a large positive effect in the public sector, the sector that receives most of the weight in

the regression of column 14.

21

The results in Table 3 also show that female candidates are roughly 1.2 percentage points

more likely to receive a positive response than male candidates in the unweighted sample, and

this estimate is significantly different from 0 at the 10% level.21 However, this effect reverses in

sign and becomes statistically insignificant in the weighted sample, suggesting that the effect is

not driven by public schools. Furthermore, candidates from a different state than the receiving

school is located in have a nearly three percentage point lower chance of receiving a positive

response than an in-state candidate does in the unweighted sample, and this result is highly

statistically significant. The magnitudes shrink slightly in the weighted sample but are still quite

large. The coefficients on the additional covariates suggest that charter schools and private

schools are more likely to give a positive response than public schools are and that schools with

a higher share of underrepresented minority students are more likely to give a positive response

than schools with a lower share are.

Table 4 shows the main results for interview requests in a regression context. There is a

concern that these estimates might lack precision due to the overall lower incidence of interview

requests than positive responses more broadly defined, but the results are qualitatively similar to

those observed in Table 3 for positive responses. However, the magnitudes in Table 4 are

generally lower than those in Table 3, which is in line with the lower mean of the left-hand side

variable in Table 4.

Table 5 explores the differences by school sector in greater detail. This table shows the

raw positive response and interview request rates for the various treatment conditions separately

for traditional public schools, charter schools, and private schools. This table shows that having

better academic qualifications is not associated with a higher likelihood of success in any of the

21 In results not shown here, I also regressed the outcome variables used in this study on dummies for the particular first and last names used in the resumes. Results from F-tests suggest no differences between individual names.

22

three sectors, with the exception that college selectivity appears to matter for public schools.

This table also shows that resumes listing male names were more likely than those listing female

names to receive a positive response or an interview request at public schools, while the opposite

was true at charter schools and private schools. All three sectors appear to show a large

preference for in-state candidates.

Table 6 presents the results by sector in a regression context. This table shows the results

of three regressions, one for each sector. These regressions include the full set of control

variables from Table 3, including the state indicator variables. The results suggest that public

schools may actually have a preference for more academically talented teachers, at least when

academic talent is measured by the selectivity level of the college the applicant attended. On the

other hand, charter and private schools do not seem to have this preference. If true, this result is

interesting because charter and private schools are often believed to make better hiring decisions

than traditional public schools due to having greater flexibility and facing more competitive

pressure. On the other hand, the coefficients on the charter school and private school indicator

variables in Tables 3 and 4 show that these schools are more likely to respond positively to

candidates than traditional public schools are. This suggests that these schools may be following

a strategy similar to the one advocated by Staiger and Rockoff (2010), in which schools have low

barriers to entry when hiring teachers but high standards for retention.

The results of Table 6 also suggest that any preference for female candidates in the

experiment is driven by private schools and perhaps also charter schools. Finally, the out-of-

state disadvantage is large and highly significant for all three sectors. Whatever the reason for

the more positive response given to in-state candidates, it must be something that affects all three

sectors. One possibility is that schools may believe that out-of-state teachers are not actually

23

interested in coming. However, the labor market for teachers is very thick and so it is not clear

why teachers would apply for jobs in places where they do not actually want to live.

Furthermore, due to the general rigidity of teacher pay, searching for outside offers in an attempt

to raise one’s pay is unlikely in this market as well. Table 7, which focuses on interview

requests, shows similar results as Table 6.

Table 8 explores whether there are differences by the level of the school. This table

shows the raw positive response and interview request rates for resumes listing various

characteristics separately for elementary schools and secondary schools. It also shows results

that reweight by enrollment in the three sectors. A notable difference between the two levels is

that higher GPAs are associated with a greater likelihood of success at elementary schools, while

the reverse is true for high school mathematics teachers. The results for college selectivity are

similar for the two levels. There appears to be a preference for females at both levels in the

unweighted results, although this is especially true at the secondary level. This preference for

females is not observed in the weighted results. The secondary level also displays a greater

preference for in-state candidates than the elementary level does, although the effects are large at

the elementary level as well. Tables 9 and 10 show the results by level in a regression context.

Each of these tables reports the results of four regressions, and each of these regressions includes

the full set of right-hand side variables from Table 3. The results in these tables support the

results of Table 8. The results are broadly similar for the two levels, although there are some

differences. For example, having a high GPA is associated with lower callbacks at the secondary

level but not the elementary level. Furthermore, the slight preference for female teachers seen in

Table 3 appears to be coming primarily from the secondary level.

24

VI. Conclusion

The results of this resume audit study suggest that an applicant’s academic credentials

have little impact on the likelihood of success at private and charter schools, although public

schools respond more favorably to candidates from more selective colleges. Additionally,

private schools demonstrate a slight preference for female candidates, and all three sectors

demonstrate a preference for in-state candidates.

This paper’s results have implications for whether offering higher salaries would result in

better teachers. To the extent that schools are hiring from the available pool suboptimally, the

potential to raise teacher quality by increasing salaries may be dampened; rather, it becomes

important what the effect of teacher salaries on the pool of applicants is. Additionally, from the

standpoint of applicants, it may be useful to compare the results of this paper to the college

quality literature. A large literature, including Black and Smith (2006), Hoekstra (2009), and

Long (2008, 2010), finds labor market returns to college quality. The results of this paper

suggest a positive effect of college quality when applying for jobs at public schools but not at

charter schools or private schools.

Because teachers have such a large impact on the life trajectories of their students,

staffing schools with the best teachers is a crucial public policy goal. The results of this paper

likely suggest some areas for optimism regarding teacher hiring practices in the United States,

but also some areas for concern. But because there is still so little that is known about teacher

hiring practices, additional research on the topic would potentially be very valuable.

25

References

Aaronson, Daniel, Lisa Barrow, and William Sander (2007), “Teachers and Student Achievement in the Chicago Public High Schools,” Journal of Labor Economics 25:1, 95-135. Altonji, Joseph G. and Rebecca M. Blank (1999), “Race and Gender in the Labor Market,” in Handbook of Labor Economics Volume 3C, Orley C. Ashenfelter and David Card, eds., Amsterdam: North-Holland, 3143-3259. Antecol, Heather, Ozkan Eren, and Serkan Ozbeklik (forthcoming), “The Effect of Teacher Gender on Student Achievement in Primary School,” Journal of Labor Economics. Anthony, Rebecca and Gerald Roe (2003), 101 Grade A Resumes for Teachers, Third Edition, Hauppage, NY: Barron’s. Bacolod, Marigee P. (2007), “Do Alternative Opportunities Matter? The Role of Female Labor Markets in the Decline of Teacher Quality,” The Review of Economics and Statistics 89:4, 737-751. Ballou, Dale (1996), “Do Public Schools Hire the Best Applicants?” The Quarterly Journal of

Economics 111:1, 97-133.

Ballou, Dale and Michael Podgursky (1997), Teacher Pay and Teacher Quality, Kalamazoo MI: W.E. Upjohn Institute for Employment Research. Balter, Dana and William D. Duncombe (2008), “Recruiting Highly Qualified Teachers: Do District Recruitment Policies Matter?” Public Finance Review 36:1, 33-62. Bertrand, Marianne and Sendhil Mullainathan (2004), “Are Emily and Greg More Employable

than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination,” The American

Economic Review 94:4, 991-1013.

Black, Dan A. and Jeffrey A. Smith (2006), “Estimating the Returns to College Quality with Multiple Proxies for Quality,” Journal of Labor Economics 24:3, 701-728. Boyd, Don, Hamp Lankford, Susanna Loeb, Matthew Ronfeldt, and Jim Wyckoff (2011), “The Role of Teacher Quality in Retention and Hiring: Using Applications to Transfer to Uncover Preferences of Teachers and Schools,” Journal of Policy Analysis and Management 30:1, 88-110. Boyd, Donald, Hamilton Lankford, Susanna Loeb, and James Wyckoff (2005a), “The Draw of Home: How Teachers’ Preferences for Proximity Disadvantage Urban Schools,” Journal of

Policy Analysis and Management 24:1, 113-132. Boyd, Donald, Hamilton Lankford, Susanna Loeb, and James Wyckoff (2005a), “Explaining the Short Careers of High-Achieving Teachers in Schools with Low-Performing Students,” The

26

American Economic Review: Papers and Proceedings of the Annual Meeting of the American

Economic Association 95:2, 166-171. Boyd, Donald, Hamilton Lankford, Susanna Loeb, and James Wyckoff (2013), “Analyzing the

Determinants of the Matching of Public School Teachers to Jobs: Disentangling the Preferences

of Teachers and Employers,” Journal of Labor Economics 31:1, 83-117.

Brause, Rita S., Christine P. Donohue, and Alice W. Ryan (2002), Succeeding at Your Interview:

A Practical Guide for Teachers, Mahwah, NJ: Lawrence Erlbaum Associates. Chetty, Raj, John N. Friedman, and Jonah E. Rockoff (2011), “The Long-Term Impacts of Teachers: Teacher Value-Added and Student Outcomes in Adulthood,” National Bureau of Economic Research Working Paper 17699. Clement, Mary C. (2007), The Definitive Guide to Getting a Teaching Job: An Insider’s Guide to

Finding the Right Job, Writing the Perfect Resume, and Nailing the Interview, Lanham, MD: Rowman & Littlefield Education. Clotfelter, Charles T., Helen F. Ladd, and Jacob L. Vigdor (2005), “Who Teaches Whom? Race and the Distribution of Novice Teachers,” Economics of Education Review 24:4, 377-392. Clotfelter, Charles T., Helen F. Ladd, and Jacob L. Vigdor (2006), “Teacher-Student Matching and the Assessment of Teacher Effectiveness,” The Journal of Human Resources 41:4, 778-820. Clotfelter, Charles T., Helen F. Ladd, and Jacob L. Vigdor (2010), “Teacher Credentials and Student Achievement in High School: A Cross-Subject Analysis with Student Fixed Effects,” The Journal of Human Resources 45:3, 655-681. Corcoran, Sean P., William N. Evans, and Robert M. Schwab (2004a), “Changing Labor-Market Opportunities for Women and the Quality of Teachers, 1957-2000,” The American Economic

Review: Papers and Proceedings of the Annual Meeting of the American Economic Association

94:2, 230-235. Corcoran, Sean P., William N. Evans, and Robert M. Schwab (2004b), “Women, the Labor Market, and the Declining Relative Quality of Teachers,” Journal of Policy Analysis and

Management 23:3, 449-470. Dee, Thomas S. (2005), “A Teacher Like Me: Does Race, Ethnicity, or Gender Matter?” The

American Economic Review: Papers and Proceedings of the Annual Meeting of the American

Economic Association 95:2, 158-165. Dee, Thomas S. (2007), “Teachers and the Gender Gaps in Student Achievement,” The Journal

of Human Resources 42:3, 528-554. Deming, David J. (2011), “Better Schools, Less Crime,” The Quarterly Journal of Economics

126:4, 2063-2115.

27

Dobbie, Will (2011), “Teacher Characteristics and Student Achievement: Evidence from Teach for America,” unpublished working paper. Enelow, Wendy S. and Louise M. Kursmark (2011), Expert Resumes for Teachers and

Educators, Third Edition, Indianapolis: JIST Works. Engel, Mimi (2012), “The Timing of Teacher Hires and Teacher Qualifications: Is There an Association?” Teachers College Record 114:12, 1-29. Engel, Mimi and Brian A. Jacob (2011), “New Evidence on Teacher Labor Supply,” National Bureau of Economic Research Working Paper 16802. Feirsen, Robert and Seth Weitzman (2004), How to Get the Teaching Job You Want: The

Complete Guide for College Graduates, Teachers Changing Schools, Returning Teachers and

Career Changers, Sterling, VA: Stylus Publishing. Glazerman, Steven, Daniel Mayer, and Paul Decker (2006), “Alternative Routes to Teaching: The Impacts of Teach for America on Student Achievement and Other Outcomes,” Journal of

Policy Analysis and Management 25:1, 75-96. Goldhaber, Dan (2007), “Everyone’s Doing It, But What Does Teacher Testing Tell Us About Teacher Effectiveness,” The Journal of Human Resources 42:4, 765-794. Goldhaber, Dan (2008), “Teachers Matter, But Effective Teacher Quality Policies are Elusive,”

in Handbook of Research in Education Finance and Policy, Helen F. Ladd and Edward B. Fiske,

eds., New York: Routledge, 146-165.

Goldhaber, Dan, Betheny Gross, and Daniel Player (2011), “Teacher Career Paths, Teacher

Quality, and Persistence in the Classroom: Are Public Schools Keeping Their Best?” Journal of

Policy Analysis and Management 30:1, 57-87.

Goldin, Claudia, Lawrence F. Katz, and Ilyana Kuziemko (2006), “The Homecoming of American College Women: The Reversal of the College Gender Gap,” The Journal of Economic

Perspectives 20:4, 133-156. Goldin, Claudia and Cecelia Rouse (2000), “Orchestrating Impartiality: The Impact of ‘Blind’ Auditions on Female Musicians,” The American Economic Review 90:4, 715-741. Gormley, William (2012), “What Our Schools Need? A Few Good Men,” USA Today August 14,

7A.

Hanushek, Eric A. (1975), “Teacher Characteristics and Gains in Student Achievement: Estimation Using Micro Data,” The American Economic Review: Papers and Proceedings of the

Annual Meeting of the American Economic Association 61:2, 280-288

28

Hanushek, Eric A. (2011), “The Economic Value of Higher Teacher Quality,” Economics of

Education Review 30:3, 466-479. Harris, Douglas N., Stacey A. Rutledge, William K. Ingle, and Cynthia C. Thompson (2010), “Mix and Match: What Principals Really Look For When Hiring Teachers,” Education Finance

and Policy 5:2, 228-246. Harris, Douglas N. and Tim R. Sass (2011), “Teacher Training, Teacher Quality and Student Achievement,” Journal of Public Economics 95:7-8, 798-812. Heckman, James J. and Peter Siegelman (1993), “The Urban Institute Audit Studies: Their Methods and Findings,” in Clear and Convincing Evidence: Measurement of Discrimination in

America, Michael Fix and Raymond J. Struyk, eds., Washington DC: The Urban Institute Press, 187-258. Hoekstra, Mark (2009), “The Effect of Attending the Flagship State University on Earnings: A Discontinuity-Based Approach,” The Review of Economics and Statistics 91:4, 717-724. Hougan, Eric (2011), Road to Teaching: A Guide to Student Teacher, Teacher Training and

Finding a Job. Hoxby, Caroline M. and Andrew Leigh (2004), “Pulled Away or Pushed Out? Explaining the Decline in Teacher Aptitude in the United States,” The American Economic Review: Papers and

Proceedings of the Annual Meeting of the American Economic Association 94:2, 236-240. Jacob, Brian A. (2002), “Where the Boys Aren’t: Non-Cognitive Skills, Returns to School and the Gender Gap in Higher Education,” Economics of Education Review 21:6, 589-598. Jackson, C. Kirabo (2009), “Student Demographics, Teacher Sorting, and Teacher Quality: Evidence from the End of School Desegregation,” Journal of Labor Economics 27:2, 213-256. Jackson, C. Kirabo and Elias Bruegmann (2009), “Teaching Students and Teaching Each Other: The Importance of Peer Learning for Teachers,” American Economic Journal: Applied

Economics 1:4, 85-108. Jepsen, Christopher and Steven Rivkin (2009), “Class Size Reduction and Student Achievement: The Potential Tradeoff between Teacher Quality and Class Size,” The Journal of Human

Resources 44:1, 223-250. Jowell, Roger and Patricia Prescott-Clarke (1970), “Racial Discrimination and White-Collar

Workers in Britain,” Race 11:4, 397-417.

Kane, Thomas J. and Douglas O. Staiger (2005), “Using Imperfect Information to Identify Effective Teachers,” unpublished working paper.

29

Killeen, Kieran M., Susanna Loeb, and Imeh Williams (2013), “A Double Draw of Proximity:

The Importance of Geography in Teacher Application and Hiring Decisions,” unpublished

working paper.

Koedel, Cory (2011), “Grading Standards in Education Departments at Universities,” Education

Policy Analysis Archives 19:23, 1-23.

Koedel, Cory and Eric Tyhurst (2012), “Math Skills and Labor-Market Outcomes: Evidence

from a Resume-Based Field Experiment,” Economics of Education Review 31:1, 131-140.

Kroft, Kory, Fabian Lange, and Matthew J. Notowidigdo (2013), “Duration Dependence and Labor Market Conditions: Evidence from a Field Experiment,” The Quarterly Journal of

Economics 128:3, 1123-1167. Lahey, Joanna N. (2008), “Age, Women, and Hiring: An Experimental Study,” The Journal of

Human Resources 43:1, 30-56.

Lankford, Hamilton, Susanna Loeb, and James Wyckoff (2002), “Teacher Sorting and the Plight

of Urban Schools: A Descriptive Analysis,” Educational Evaluation and Policy Analysis 24:1,

37-62.

Levin, Jessica and Meredith Quinn (2003), Missed Opportunities: How We Keep High-Quality

Teachers Out of Urban Classrooms, New York: The New Teacher Project.

Lochner, Lance and Enrico Moretti (2004), “The Effect of Education on Crime: Evidence from Prison Inmates, Arrests, and Self-Reports,” The American Economic Review 94:1, 155-189. Long, Mark C. (2008), “College Quality and Early Adult Outcomes,” Economics of Education

Review 27:5, 588-602. Long, Mark C. (2010), “Changes in the Returns to Education and College Quality,” Economics

of Education Review 29:3, 338-347. McKinney, Anne (2000), Real –Resumes for Teachers, Fayetteville, NC: PREP Publishing. Murnane, Richard (1975), The Impact of School Resources on the Learning of Inner City

Children, Cambridge MA: Ballinger. Neumark, David, Roy J. Bank, and Kyle D. Van Nort (1996), “Sex Discrimination in Restaurant Hiring: An Audit Study,” The Quarterly Journal of Economics 111:3, 915-941. Neumark, David (2012), “Detecting Discrimination in Audit and Correspondence Studies,” The

Journal of Human Resources 47:4, 1128-1157.

30

Ondrich, Jan, Stephen Ross, and John Yinger (2003), “Now You See It, Now You Don't: Why Do Real Estate Agents Withhold Available Houses from Black Customers?” The Review of

Economics and Statistics 85:4, 854-873. Oreopoulos, Philip (2011), “Why Do Skilled Immigrants Struggle in the Labor Market? A Field Experiment with Thirteen Thousand Resumes,” American Economic Journal: Economic Policy

3:4, 148-171. Papay, John P., Matthew A. Kraft, Julia Bloom, Kate Buckley, and David Liebowitz (2013), “Missed Opportunities in the Labor Market or Temporary Disruptions? How Late Teacher Hiring Affects Student Achievement,” unpublished working paper. Pollock, Robert (2011), Teacher Interviews: How to Get Them & How to Get Hired! 2

nd Edition,

Knoxville, TN: Advanta Publishing. Ransom, Michael R. and David P. Sims (2010), “Estimating the Firm’s Labor Supply Curve in a ‘New Monopsony’ Framework: Schoolteachers in Missouri,” Journal of Labor Economics 28:2, 331-355. Reininger, Michelle (2012), “Hometown Disadvantage? It Depends on Where You’re From: Teachers’ Location Preferences and the Implications for Staffing Schools,” Educational

Evaluation and Policy Analysis 34:2, 127-145. Riach, Peter A. and Judith Rich (2006), “An Experimental Investigation of Sexual Discrimination in Hiring in the English Labor Market,” Advances in Economic Analysis & Policy

6:2, Article 1, 1-20. Rivkin, Steven G., Eric A. Hanushek, and John F. Kain (2005), “Teachers, Schools, and Academic Achievement,” Econometrica 73:2, 417-459. Rockoff, Jonah E. (2004), “The Impact of Individual Teachers on Student Achievement: Evidence from Panel Data,” The American Economic Review: Papers and Proceedings of the

Annual Meeting of the American Economic Association 94:2, 247-252. Rooth, Dan-Olof (2009), “Obesity, Attractiveness, and Differential Treatment in Hiring: A Field Experiment,” The Journal of Human Resources 44:3, 710-735. Rothstein, Jesse (2009), “Student Sorting and Bias in Value-Added Estimation: Selection on Observables and Unobservables,” Education Finance and Policy 4:4, 537-571. Rothstein, Jesse (2010), “Teacher Quality in Educational Production: Tracking, Decay, and Student Achievement,” The Quarterly Journal of Economics 125:1, 175-214. Rothstein, Jesse (2012), “Teacher Quality Policy When Supply Matters,” unpublished working paper.

31

Rutledge, Stacey A., Douglas N. Harris, Cynthia T. Thompson, and W. Kyle Ingle (2008), “Certify, Blink, Hire: An Examination of the Process and Tools of Teacher Selection and Screening,” Leadership and Policy in Schools 7:3, 237-263. Staiger, Douglas O. and Jonah E. Rockoff (2010), “Searching for Effective Teachers with Imperfect Information,” The Journal of Economic Perspectives 24:3, 97-117. Strauss, Robert P., Lori R. Bowes, Mindy S. Marks, and Mark R. Plesko (2000), “Improving

Teacher Preparation and Selection: Lessons from the Pennsylvania Experience,” Economics of

Education Review 19:4, 387-415.

Warner, Jack, Clyde Bryan, and Diane Warner (2006), Inside Secrets of Finding a Teaching Job:

The Most Effective Search Methods for Both New and Experienced Educators, Third Edition,

Indianapolis: JIST Works.

Wei, Tim (2010), I Want a Teaching Job! Guide to Getting the Teaching Job of Your Dreams,

TPW Websites.

Weichselbaumer, Doris (2003), “Sexual Orientation Discrimination in Hiring,” Labour

Economics 10:6, 629-642.

Xu, Zeyu, Jane Hannaway, and Colin Taylor (2011), “Making a Difference? The Effects of

Teach For America in High School,” Journal of Policy Analysis and Management 30:3, 447-469.

Yinger, John (1986), “Measuring Discrimination with Fair Housing Audits: Caught in the Act,”

The American Economic Review 76:5, 881-893.

32

Unweighted Weighted

Variable Mean Mean

Outcomes

Positive Response 0.0787 0.0514

Interview Request 0.0433 0.0315

Resume Characteristics

High GPA 0.3255 0.3173

Medium GPA 0.3438 0.3440

Low GPA 0.3307 0.3387

High College Selectivity 0.3298 0.3255

Medium College Selectivity 0.3442 0.3491

Low College Selectivity 0.3260 0.3253

Female 0.5000 0.5038

Out-of-State 0.2565 0.2637

Science 0.1200 0.1045

School Characteristics

Secondary School 0.4663 0.4218

Charter School 0.3333 0.0301

Private School 0.3333 0.0807

Traditional Public School 0.3333 0.8892

Fraction Underrepresented Minority 0.3479 0.3499

Located in City 0.3937 0.2495

Located in Suburb 0.3007 0.2971

Located in Town 0.1007 0.1318

Located in Rural Area 0.2050 0.3216

Table 1: Summary Statistics

Notes: The sample size is 6,000. The unweighted standard deviation of the "Fraction Underrepresented Minority" variable is 0.3619, and the weighted standard deviation is 0.3348. All other variables are binary.

33

Positive Interview Positive Interview

High GPA 0.074 0.039 0.049 0.028

Medium GPA 0.079 0.045 0.051 0.033

Low GPA 0.083 0.045 0.054 0.033

High College Selectivity 0.084 0.045 0.067 0.036

Medium College Selectivity 0.079 0.043 0.051 0.034

Low College Selectivity 0.073 0.041 0.036 0.024

Male 0.073 0.039 0.053 0.031

Female 0.085 0.047 0.049 0.031

In-State 0.086 0.049 0.057 0.036

Out-of-State 0.057 0.027 0.035 0.018

Unweighted Weighted

Table 2: Raw Positive Response and Interview Request Rates

34

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)

High GPA -0.0084 -0.0084 -0.0098 -0.0083 -0.0056 -0.0062 -0.0083 -0.0084

(0.0086) (0.0086) (0.0085) (0.0085) (0.0103) (0.0103) (0.0102) (0.0099)

Medium GPA -0.0037 -0.0025 -0.0021 -0.0007 -0.0033 -0.0026 -0.0008 -0.0031

(0.0084) (0.0084) (0.0083) (0.0082) (0.0100) (0.0099) (0.0098) (0.0094)

High College Selectivity 0.0113 0.0119 0.0121 0.0129 0.0306*** 0.0310*** 0.0337*** 0.0361***

(0.0084) (0.0084) (0.0084) (0.0084) (0.0099) (0.0099) (0.0098) (0.0101)

Medium College Selectivity 0.0068 0.0072 0.0080 0.0082 0.0144 0.0150* 0.0166* 0.0184**

(0.0083) (0.0083) (0.0082) (0.0082) (0.0090) (0.0090) (0.0091) (0.0091)

Female 0.0120* 0.0122* 0.0125* 0.0119* -0.0039 -0.0040 -0.0042 -0.0064

(0.0070) (0.0070) (0.0069) (0.0070) (0.0080) (0.0080) (0.0079) (0.0081)

Out-of-State -0.0289***-0.0293***-0.0289***-0.0269*** -0.0220** -0.0221**-0.0224***-0.0231***

(0.0072) (0.0072) (0.0072) (0.0072) (0.0086) (0.0086) (0.0086) (0.0087)

Science 0.0216* 0.0208 0.0329* 0.0348*

(0.0131) (0.0130) (0.0183) (0.0181)

Secondary School -0.0024 -0.0022 -0.0012 0.0016

(0.0083) (0.0083) (0.0102) (0.0097)

Charter School 0.0555*** 0.0517*** 0.0541*** 0.0533***

(0.0101) (0.0106) (0.0109) (0.0119)

Private School 0.0376*** 0.0423*** 0.0481*** 0.0507***

(0.0094) (0.0096) (0.0109) (0.0110)

Fraction Underrepresented Minority 0.0317** 0.0330** 0.0571*** 0.0539**

(0.0150) (0.0159) (0.0213) (0.0211)

Urbanicity Dummies? No No No No No Yes Yes No No No No No Yes Yes

State Dummies? No No No No No No Yes No No No No No No Yes

N 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000

Unweighted Weighted

Table 3: Determinants of Positive Responses

Notes: Standard errors that are robust to clustering at the school level are in parentheses. A single asterisk denotes significance at the 10% level, a double asterisk denotes significance at the 5% level, and a triple asterisk denotes statistical significance at the 1% level.

35

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)

High GPA -0.0059 -0.0059 -0.0069 -0.0066 -0.0052 -0.0052 -0.0067 -0.0082

(0.0064) (0.0063) (0.0063) (0.0063) (0.0084) (0.0083) (0.0083) (0.0079)

Medium GPA -0.0003 0.0005 0.0019 0.0023 -0.0000 0.0005 0.0031 0.0001

(0.0065) (0.0064) (0.0063) (0.0062) (0.0084) (0.0084) (0.0083) (0.0079)

High College Selectivity 0.0041 0.0046 0.0054 0.0059 0.0117 0.0121 0.0149* 0.0159**

(0.0064) (0.0064) (0.0063) (0.0063) (0.0079) (0.0079) (0.0079) (0.0080)

Medium College Selectivity 0.0017 0.0020 0.0031 0.0030 0.0091 0.0097 0.0109 0.0108

(0.0062) (0.0062) (0.0061) (0.0061) (0.0078) (0.0078) (0.0079) (0.0077)

Female 0.0080 0.0082 0.0087* 0.0081 0.0000 0.0002 0.0001 -0.0017

(0.0053) (0.0052) (0.0052) (0.0052) (0.0062) (0.0062) (0.0061) (0.0062)

Out-of-State -0.0216***-0.0219***-0.0218***-0.0205*** -0.0184***-0.0185***-0.0191***-0.0194***

(0.0053) (0.0053) (0.0053) (0.0053) (0.0068) (0.0068) (0.0067) (0.0067)

Science 0.0165 0.0167 0.0198 0.0227

(0.0110) (0.0109) (0.0153) (0.0151)

Secondary School 0.0149** 0.0147** 0.0098 0.0105

(0.0063) (0.0063) (0.0084) (0.0080)

Charter School 0.0312*** 0.0253*** 0.0293*** 0.0290***

(0.0082) (0.0086) (0.0093) (0.0103)

Private School 0.0106 0.0121* 0.0131 0.0112

(0.0066) (0.0066) (0.0080) (0.0082)

Fraction Underrepresented Minority 0.0497*** 0.0479*** 0.0568*** 0.0462***

(0.0118) (0.0122) (0.0184) (0.0175)

Urbanicity Dummies? No No No No No Yes Yes No No No No No Yes Yes

State Dummies? No No No No No No Yes No No No No No No Yes

N 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000 6,000

Table 4: Determinants of Interview Requests

Unweighted Weighted


36

Public Charter Private Public Charter Private

High GPA 0.044 0.101 0.076 0.027 0.065 0.025

Medium GPA 0.047 0.114 0.077 0.032 0.073 0.030

Low GPA 0.050 0.121 0.079 0.032 0.081 0.024

High College Selectivity 0.065 0.115 0.073 0.035 0.071 0.031

Medium College Selectivity 0.046 0.104 0.090 0.033 0.066 0.030

Low College Selectivity 0.031 0.118 0.068 0.023 0.082 0.019

Male 0.051 0.101 0.067 0.031 0.068 0.020

Female 0.044 0.123 0.088 0.030 0.078 0.034

In-State 0.052 0.119 0.086 0.035 0.081 0.030

Out-of-State 0.032 0.090 0.051 0.017 0.048 0.018

Positive Interview

Table 5: Raw Positive Response and Interview Request Rates by Sector

37

Variable Public Charter Private

High GPA -0.0092 -0.0159 -0.0028

(0.0112) (0.0181) (0.0139)

Medium GPA -0.0045 -0.0030 -0.0043

(0.0107) (0.0175) (0.0140)

High College Selectivity 0.0409*** -0.0059 0.0017

(0.0115) (0.0177) (0.0144)

Medium College Selectivity 0.0197* -0.0147 0.0160

(0.0103) (0.0170) (0.0147)

Female -0.0109 0.0188 0.0215*

(0.0091) (0.0143) (0.0118)

Out-of-State -0.0221** -0.0279* -0.0309**

(0.0098) (0.0158) (0.0120)

Science 0.0398* 0.0212 0.0036

(0.0210) (0.0239) (0.0216)

Secondary School 0.0029 0.0028 0.0013

(0.0111) (0.0173) (0.0155)

Fraction Underrepresented Minority 0.0604** 0.0339 -0.0244

(0.0238) (0.0297) (0.0407)

Urbanicity Dummies? Yes Yes Yes

State Dummies? Yes Yes Yes

N 2,000 2,000 2,000

Table 6: Determinants of Positive Responses by Sector


38

Variable Public Charter Private

High GPA -0.0092 -0.0113 0.0004

(0.0090) (0.0143) (0.0083)

Medium GPA -0.0012 -0.0020 0.0071

(0.0090) (0.0142) (0.0086)

High College Selectivity 0.0175* -0.0127 0.0132

(0.0092) (0.0143) (0.0089)

Medium College Selectivity 0.0118 -0.0160 0.0109

(0.0087) (0.0139) (0.0081)

Female -0.0043 0.0096 0.0153*

(0.0071) (0.0116) (0.0078)

Out-of-State -0.0201*** -0.0318*** -0.0119

(0.0076) (0.0123) (0.0073)

Science 0.0243 0.0087 0.0192

(0.0175) (0.0212) (0.0169)

Secondary School 0.0101 0.0235* 0.0192**

(0.0091) (0.0141) (0.0096)

Fraction Underrepresented Minority 0.0462** 0.0546** 0.0138

(0.0199) (0.0248) (0.0210)

Urbanicity Dummies? Yes Yes Yes

State Dummies? Yes Yes Yes

N 2,000 2,000 2,000

Table 7: Determinants of Interview Requests by Sector


39

Elementary Secondary Elementary Secondary Elementary Secondary Elementary Secondary

High GPA 0.078 0.070 0.035 0.044 0.053 0.043 0.027 0.029

Medium GPA 0.075 0.084 0.036 0.056 0.046 0.059 0.023 0.048

Low GPA 0.072 0.094 0.027 0.065 0.043 0.069 0.023 0.047

High College Selectivity 0.079 0.089 0.032 0.061 0.063 0.074 0.028 0.049

Medium College Selectivity 0.075 0.084 0.033 0.056 0.048 0.055 0.028 0.041

Low College Selectivity 0.070 0.076 0.033 0.050 0.030 0.045 0.017 0.034

Male 0.071 0.074 0.030 0.050 0.050 0.057 0.024 0.041

Female 0.078 0.092 0.036 0.061 0.044 0.057 0.024 0.042

In-State 0.081 0.091 0.036 0.064 0.050 0.066 0.026 0.050

Out-of-State 0.058 0.057 0.025 0.031 0.039 0.029 0.021 0.013

Table 8: Raw Positive Response and Interview Request Rates by Level

Positive Interview

Unweighted Weighted

Positive Interview

40

Variable Elementary Secondary Elementary Secondary

High GPA 0.0047 -0.0251** 0.0060 -0.0288*

(0.0116) (0.0125) (0.0123) (0.0165)

Medium GPA 0.0034 -0.0053 0.0005 -0.0047

(0.0110) (0.0126) (0.0125) (0.0152)

High College Selectivity 0.0124 0.0152 0.0378*** 0.0298*

(0.0117) (0.0123) (0.0124) (0.0173)

Medium College Selectivity 0.0058 0.0111 0.0215** 0.0114

(0.0111) (0.0126) (0.0108) (0.0150)

Female 0.0056 0.0177* -0.0121 0.0020

(0.0092) (0.0106) (0.0104) (0.0124)

Out-of-State -0.0200** -0.0364*** -0.0107 -0.0397***

(0.0096) (0.0111) (0.0111) (0.0145)

Science 0.0223* 0.0378**

(0.0131) (0.0184)

Charter School 0.0592*** 0.0451*** 0.0567*** 0.0437**

(0.0150) (0.0159) (0.0163) (0.0185)

Private School 0.0590*** 0.0275* 0.0818*** 0.0229

(0.0130) (0.0145) (0.0144) (0.0161)

Fraction Underrepresented Minority 0.0440* 0.0251 0.0945*** 0.0156

(0.0227) (0.0228) (0.0303) (0.0275)

Urbanicity Dummies? Yes Yes Yes Yes

State Dummies? Yes Yes Yes Yes

N 3,202 2,798 3,202 2,798

Table 9: Determinants of Positive Responses by Level

Unweighted Weighted


41

Variable Elementary Secondary Elementary Secondary

High GPA 0.0068 -0.0217** -0.0008 -0.0191

(0.0074) (0.0103) (0.0087) (0.0141)

Medium GPA 0.0084 -0.0051 -0.0012 0.0051

(0.0072) (0.0108) (0.0098) (0.0138)

High College Selectivity 0.0013 0.0135 0.0145 0.0159

(0.0078) (0.0103) (0.0089) (0.0152)

Medium College Selectivity 0.0007 0.0074 0.0132 0.0086

(0.0074) (0.0102) (0.0086) (0.0135)

Female 0.0048 0.0110 -0.0041 0.0040

(0.0062) (0.0087) (0.0073) (0.0103)

Out-of-State -0.0099 -0.0349*** -0.0052 -0.0412***

(0.0065) (0.0086) (0.0083) (0.0112)

Science 0.0178 0.0252

(0.0110) (0.0154)

Charter School 0.0253** 0.0286** 0.0235* 0.0318**

(0.0112) (0.0135) (0.0134) (0.0156)

Private School 0.0180** 0.0096 0.0298*** -0.0006

(0.0073) (0.0116) (0.0094) (0.0130)

Fraction Underrepresented Minority 0.0557*** 0.0389** 0.0841*** 0.0117

(0.0160) (0.0187) (0.0254) (0.0229)

Urbanicity Dummies? Yes Yes Yes Yes

State Dummies? Yes Yes Yes Yes

N 3,202 2,798 3,202 2,798

Table 10: Determinants of Interview Requests by Level

Unweighted Weighted


what kind of teachers are schools looking for? evidence ... · rothstein, john yinger; seminar...

Documents